Knowledge in my view is a form of action. It involves endeavors to get it right, and more broadly it concerns aimings, which can be functional rather than intentional.

The risk pertinent to a particular attempt (and to its evaluation as an attempt of its sort) is the risk that the agent will fail to attain the end constitutive of that attempt. This risk of failure is coordinate with how likely or unlikely it may be that the agent will then succeed.

In arriving at the relevant theory about the specifics of our faculty of vision we will presumably use our eyes to gather relevant data. Based on such data we come to know about the optic nerve, the structure of our eyes, the rods and cones, etc., so as to explain how it is that vision gives us reliable access to the shapes and colors of objects around us. In reliably arriving at that theory we thus exercise the very faculty whose reliability is explained by the theory. There is no vice in this sort of circularity.

Normal adults can doodle, amble, and drift with no need to assess risk, since there is normally no risk at all. Jazz improvisation seems less subject to standards of risk than surgery, and less than much formal athletic performance, as in a tennis match.

In my view (animal) knowledge is apt belief, where not only the belief (its existence and content) but also its correctness is creditable to the subject's competence.

Epistemic competence might be posterior to knowledge conceptually, however, while still prior metaphysically.

If the agent aims to make the attempt if and only if it would be apt, then a distinctive element of risk assessment becomes relevant: How probably would the agent succeed in attempting that fuller end?